pith. sign in

arxiv: 1907.06191 · v1 · pith:HTUWJXLWnew · submitted 2019-07-14 · 🧮 math.NA · cs.NA· physics.comp-ph

A GPU implementation of the Discontinuous Galerkin method for simulation of diffusion in brain tissue

Daniel Cervantes , Miguel Angel Moreles , Joaquin Pe\~na , Alonso Ramirez-Manzanares This is my paper

Pith reviewed 2026-05-24 21:50 UTC · model grok-4.3

classification 🧮 math.NA cs.NAphysics.comp-ph
keywords Discontinuous GalerkinGPUCUDAdiffusion equationbrain tissuecovariance matrixnumerical simulationparallel computing
0
0 comments X

The pith

A CUDA implementation of the Discontinuous Galerkin method computes the covariance matrix for water diffusion simulations in brain tissue.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a parallel GPU code to solve the diffusion equation and extract the associated covariance matrix for modeling water movement inside brain tissue. The discretization follows the Discontinuous Galerkin approach and is written in CUDA to run on graphics processors. Results are shown only for two-dimensional test problems. The stated goal is a numerical scheme that stays consistent with the underlying physical diffusion process.

Core claim

We develop a methodology to approximate the covariance matrix associated to the simulation of water diffusion inside the brain tissue. The computation is based on an implementation of the Discontinuous Galerkin method of the diffusion equation, in accord with the physical phenomenon. The implementation is in parallel using GPUs in the CUDA language. Numerical results are presented in 2D problems.

What carries the argument

Discontinuous Galerkin discretization of the diffusion equation, executed in parallel on GPUs via CUDA.

If this is right

  • The GPU code produces covariance matrices for two-dimensional diffusion problems.
  • The parallel implementation targets faster execution of diffusion simulations compared with serial CPU versions.
  • The scheme is constructed to remain consistent with the physical diffusion process.
  • Numerical experiments are restricted to two spatial dimensions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Extending the same CUDA framework to three-dimensional domains would require only changes to the mesh and data structures.
  • The covariance output could serve as input to statistical models that estimate tissue parameters from diffusion-weighted images.
  • Similar GPU Discontinuous Galerkin codes might apply to other linear diffusion problems outside brain imaging.

Load-bearing premise

The standard diffusion equation accurately represents the physical phenomenon of water diffusion inside brain tissue.

What would settle it

Direct comparison of the computed covariance matrix against measured diffusion tensors obtained from real brain tissue samples would test whether the numerical results match experimental observations.

Figures

Figures reproduced from arXiv: 1907.06191 by Alonso Ramirez-Manzanares, Daniel Cervantes, Joaquin Pe\~na, Miguel Angel Moreles.

Figure 1
Figure 1. Figure 1: The substrate consists of 1901 non-overlapping circles. e- n￾ [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2 [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Meshing. 3.2. Numerical flux. There is no preferred direction of propagation in the heat equation, thus for u a central flux is considered, namely hu,K− (u −, u+, n −) = u − + u + 2 n −. A physical assumption is that there is no flow between axons and the extracel￾lular region. Consequently, for q, we propose the numerical flow hq,K− (q −, q +, n −) = 2k −k + k − + k + 1 2 (q − + q +) · n −. This is coined… view at source ↗
Figure 4
Figure 4. Figure 4: Left: 2D free diffusion. Right: 1D view. Let us compare the free diffusion (without axons) approximation, versus the analytical solution. The latter is a bivariate Gaussian function f(x, y; Σ) defined in (4.1), where the covariance matrix is Σ =  2T k 0 0 2T k . Taking k = 450µm/s2 and T = 0.036s, non zero coefficients in the covariance matrix are equal to 32.4. The DG-CUDA Gaussian matrix fit for 400 × … view at source ↗
Figure 5
Figure 5. Figure 5: Left: One PDE solution. Right: Normalized and cen￾tered solution for one PDE [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Left: 2D free diffusion. Right: 1D view. In both cases the approximation of the Gaussian function is highly accurate. The least squares residual (9) is of the order O(10−8 ). For practical purposes, the Gaussian density functions coincide. But the DG-CUDA approximation is structurally more consistent. The matrix is symmetric, the values in the diagonal coincide and the other terms are near zero. Case study… view at source ↗
read the original abstract

In this work we develop a methodology to approximate the covariance matrix associated to the simulation of water diffusion inside the brain tissue. The computation is based on an implementation of the Discontinuous Galerkin method of the diffusion equation, in accord with the physical phenomenon. The implementation in in parallel using GPUs in the CUDA language. Numerical results are presented in 2D problems.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The manuscript claims to develop a methodology to approximate the covariance matrix for water diffusion simulations in brain tissue. The approach relies on a CUDA GPU-parallel implementation of the Discontinuous Galerkin discretization of the diffusion equation, with numerical results shown for 2D problems.

Significance. If the implementation and results hold, the work would demonstrate a practical GPU-accelerated DG solver for a standard diffusion model in a neuroscience application, potentially enabling faster covariance computations that are relevant to diffusion MRI or tissue modeling.

major comments (1)
  1. [Abstract] Abstract: the statement that 'Numerical results are presented in 2D problems' supplies no validation data, analytic comparisons, error estimates, or convergence rates. In a numerical analysis manuscript this omission leaves the central claim about the GPU DG implementation unsupported.
minor comments (1)
  1. [Abstract] Abstract: typographical error 'The implementation in in parallel' should be 'The implementation is in parallel'.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the single major comment below.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the statement that 'Numerical results are presented in 2D problems' supplies no validation data, analytic comparisons, error estimates, or convergence rates. In a numerical analysis manuscript this omission leaves the central claim about the GPU DG implementation unsupported.

    Authors: We agree that the abstract is too brief and does not reference the specific validation elements. The manuscript presents numerical results for 2D diffusion problems that include comparisons to analytic solutions, computed error estimates, and observed convergence behavior under DG refinement. We will revise the abstract to summarize these aspects (e.g., expanding the final sentence to note validation, error estimates, and convergence rates). This change will be made in the revised version. revision: yes

Circularity Check

0 steps flagged

No significant circularity; standard numerical implementation of DG method

full rationale

The paper describes a GPU/CUDA implementation of the standard Discontinuous Galerkin discretization of the diffusion equation to generate covariance matrices from 2D simulations of brain tissue water diffusion. No equations, parameters, or results are obtained by fitting to the target outputs, self-definition, or load-bearing self-citation chains. The physical model choice (standard diffusion PDE) is an external modeling assumption, not an internal reduction. The derivation chain consists of applying a well-known numerical scheme to a standard PDE and reporting parallel performance and sample results; this is self-contained against external benchmarks and contains no circular steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the applicability of the diffusion equation to brain tissue and the standard mathematical properties of the discontinuous Galerkin discretization; no free parameters, invented entities, or additional axioms are identifiable from the abstract.

axioms (1)
  • domain assumption The diffusion equation accurately models water diffusion inside brain tissue.
    Explicitly invoked in the abstract as the basis for the computation 'in accord with the physical phenomenon'.

pith-pipeline@v0.9.0 · 5594 in / 1289 out tokens · 35839 ms · 2026-05-24T21:50:54.195935+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

  1. [1]

    Aboitiz, A.B

    F. Aboitiz, A.B. Scheibel, R.S.Fisher, and E.Zaidel. Fiber composition of the human corpus callosum. Brain Res., 143(53), 1992

    work page 1992

  2. [2]

    Alexander

    D.C. Alexander. An introduction to computational diffusion mri: the diffusion tensor and beyond. In Hagen H. Weickert J., editor, Visualization and Processing of Tensor Fields , chapter 10, pages 83–106. Springer Verlag, 2006

    work page 2006

  3. [3]

    Alexander

    D.C. Alexander. Modelling, fitting and sampling in diffusion mri. Magnetic Resonance , 103:255–260, 2009

    work page 2009

  4. [4]

    L. M. Burcaw, E. Fieremans, and D. S. Novikov. Mesoscopic structure of neuronal tracts from time-dependent diffusion. NeuroImage, 114:18–37, 2015

    work page 2015

  5. [5]

    Discontinuous Galerkin methods: theory, computation and applications , volume 11

    Bernardo Cockburn, George E Karniadakis, and Chi-Wang Shu. Discontinuous Galerkin methods: theory, computation and applications , volume 11. Springer Science & Business Media, 2012

    work page 2012

  6. [6]

    Contrast and stability of the axon diameter index from microstructure imaging with diffusion MRI

    Tim B Dyrby, Matt G Hall, Maurice Ptito, Daniel Alexander, et al. Contrast and stability of the axon diameter index from microstructure imaging with diffusion MRI. Magnetic Resonance in Medicine, 70(3):711–721, 2013

    work page 2013

  7. [7]

    Nedjati-Gilani Gemma L. et. al. Machine learning based compartment models with perme- ability for white matter microstructure imaging. Neuroimage, 150:119?135, 2017. 12 DANIEL CER V ANTES, MIGUEL ANGEL MORELES, JOAQUIN PE˜NA, AND ALONSO RAMIREZ-MANZANARES

    work page 2017

  8. [8]

    F., Stait-Gardner T., Ghadirian B., Yadav N

    Moroney B. F., Stait-Gardner T., Ghadirian B., Yadav N. N., and Price WS. Numerical analysis of nmr diffusion measurements in the short gradient pulse limit.Journal of Magnetic Resonance, 234:165–175, 2013

    work page 2013

  9. [9]

    Wald, Hui Zhang, Claudia A.M

    Uran Ferizi, Torben Schneider, Thomas Witzel, Lawrence L. Wald, Hui Zhang, Claudia A.M. Wheeler-Kingshott, and Daniel C. Alexander. White matter compartment models for in vivo diffusion MRI at 300 mt/m. NeuroImage, 118:468–483, 2015

    work page 2015

  10. [10]

    Novikov, Jens H

    Els Fieremans, Dmitry S. Novikov, Jens H. Jensen, and Joseph A. Helpern. Monte carlo study of a two-compartment exchange model of diffusion. NMR in biomedicine, 23 7:711–24, 2010

    work page 2010

  11. [11]

    Malcolm, and Glyn Johnson

    Nima Gilani, Paul N. Malcolm, and Glyn Johnson. An improved model for prostate diffusion incorporating the results of monte carlo simulations of diffusion in the cellular compartment. NMR in biomedicine , 30 12, 2017

    work page 2017

  12. [12]

    Hall and D.C

    M.G. Hall and D.C. Alexander. Convergence and parameter choice for monte-carlo simula- tions of diffusion mri. IEEE TRANSACTIONS ON MEDICAL IMAGING , 28(9), 2009

    work page 2009

  13. [13]

    Diffusion MRI theory methods and applications

    Derek Jones. Diffusion MRI theory methods and applications. Oxford University Press, 2011

    work page 2011

  14. [14]

    C.H. Neuman. Spin echo of spins diffusing in a bounded medium. Chemical Physics , 60(11):4508?4511, 1974

    work page 1974

  15. [15]

    A finite elements method to solve the bloch-torrey equation applied to diffusion magnetic resonance imaging

    Dang Van Nguyen, Jing-Rebecca Li, Denis Grebenkov, and Denis Le Bihan. A finite elements method to solve the bloch-torrey equation applied to diffusion magnetic resonance imaging. J. Comput. Physics , 263:283–302, 2014

    work page 2014

  16. [16]

    Panagiotaki, T

    E. Panagiotaki, T. Schneider, B. Siow, M.G. Hall, M.F. Lythgoe, and D.C. Alexander. Compartment models of the diffusion mr signal in brain white matter: A taxonomy and comparison. NeuroImage, 2012

    work page 2012

  17. [17]

    Diffusion tensor MR imaging of the human brain

    C Pierpaoli, P Jezzard, P J Basser, A Barnett, and G Di Chiro. Diffusion tensor MR imaging of the human brain. Radiology, 201(3):637–648, 1996

    work page 1996

  18. [18]

    W.S. Price. Pulsed-field gradient nuclear magnetic resonance as a tool for studying transla- tional diffusion: Part 1. basic theory. G-Animal’s Journal, 1997

    work page 1997

  19. [19]

    Quantifying diameter overestimation of undulating axons from synthetic dw-mri

    Alonso Ramirez-Manzanares, Mario Ocampo-Pineda, Jonathan Rafael-Patino, Giorgio Inno- centi, Jean-Philippe Thiran, and Alessandro Daducci. Quantifying diameter overestimation of undulating axons from synthetic dw-mri. InProcc. In Annual Meeting of the International Society of Magnetic Resonance in Medicine , page 748. Annual Meeting of the SMRM, Pars, Fra...

    work page 2018

  20. [20]

    van Gelderen, D

    P. van Gelderen, D. DesPers, C.M. van Zijl, and C.T.W. Moonen. Evaluation of restricted diffusion in cylinders. phosphocreatine in rabbit leg muscle. Magnetic Resonance, 103:255– 260, 1994. D. Cervantes, INFOTEC, Av. San Fernando 37, Col. Toriello Guerra, Tlalpan, Ciudad de Mexico 14050, Mexico, M. A. Moreles, J. Pe˜na, A. Ramirez-Manzanares, E-mail addres...

    work page 1994